Jake Elwes

Use an image captioning model that translates CNN features to RNN features, and another GAN that translates RNN features to images in a loop